一种基于CS-SVM的老年人异常步态识别系统

Abnormal Gait Recognition System for the Elderly Based on CS-SVM

针对老年人异常步态识别问题,提出了一种基于代价敏感支持向量机的步态识别系统.该系统首先对采集到的步态信号使用巴特沃斯带通滤波器进行滤波并使用双树复小波进行步态信号时频特征提取;其次,构造代价敏感支持向量机分类器,以提取的步态信号特征为输入对分类器进行训练;最后,对训练得到的代价敏感支持向量机分类器进行测试.实验结果表明,该分类器能够成功识别跛行、踮脚、震颤、正常四种类型的步态,平均识别率达到95%以上,而普通支持向量机识别率仅为80%左右,可见代价敏感支持向量机分类器的异常步态识别效果要优于普通支持向量机分类器的识别效果,能够实现老年人异常步态的识别,并具有准确性及可靠性.

Aiming at the problem of gait recognition in the elderly,a gait recognition system based on cost-sensitive support vector machine is proposed.The system first uses the Butterworth bandpass filter to filter the acquired gait signals and uses the double-tree complex wavelet to perform time-frequency feature extraction of gait signals.Secondly,construct the cost-sensitive support vector machine classifier,and train the classifier with the extracted gait signal characteristics as input.Finally,the trained classifier is tested.The experimental results show that the classifier can successfully identify four types of gait,such as limp,lameness,tremor and normal,and the average recognition rate is over 95%,while the recognition rate of ordinary support vector machine is only about 80%.It can be seen that the abnormal gait recognition effect of the cost-sensitive support vector machine classifier is better than that of the ordinary support vector machine classifier,and it can realize the abnormal gait recognition of the elderly,and has accuracy and reliability.